Saturday, December 19, 2009

Histograms Part 3

Read Part 1 and Part 2 first.

So today we actually get to it. Hopefully, after the last two posts, this principle has become clear: the sensor in a DSLR can only record a limited range of brightness levels in any given scene. Because of that, it would be nice to know how well we are utilizing the different brightness levels that our sensor is able to record. Our effectiveness in using our sensor in this way is all about proper exposure. And the best tool we have to know how well we have exposed a picture is a histogram. So here we go…

As I said in Part 1, a histogram is just a graph that plots data. In digital photography, the data that it plots has to do with brightness levels. The graph basically tells us how many pixels were recorded at each available brightness level. So here's what a histogram looks like:

What you are looking at is a graph with all the different brightness levels the camera could record across the bottom and the number of pixels recorded at each level up the side. Here is the image that this histogram came from:

In this image the available brightness levels are used quite well - this is a well exposed image. The hump on the left of the histogram represents most of the wood and hair while the hump on the right of the histogram represents most of the skin and dress. Here is a color-coded comparison of where the brightness levels are actually distributed:

So you can see the large blue hump represents lots of pixels in the image. Similarly, the green/yellow hump represents a lot of pixels too, but fewer than the blue hump. Obviously there aren't a lot of pixels in the red area, which is why that part of the graph is so low.

When you look at a histogram, a well exposed image will have most of the pixels distributed with the humps away from the edges. If there are humps in the graph up against either edge, you will be losing information. Consider these two pictures from  Part 1 with their corresponding histograms:

Notice how this histogram has a hump on the left representing the blackest parts of the pipe. A wide range of brightness levels that fell below the recordable range are getting assigned the lowest brightness levels. In other words, parts of the pipe that should have been recorded at levels like 0, -1, -2… were all recorded at level 1 since that is the darkest level available. The result is that there is no separation between what differences in brightness there were in real life. So instead of seeing texture in those very dark parts of the pipe, we just see black patches.

In this histogram, notice the spike on the very right edge. This spike represents every pixel in the sky of the image. Even though there were differences in the level of brightness in the sky in real life, they all fell above the range that was recorded and therefore were all assigned the highest brightness level. So in this image, let's say the brightest level is ten. The problem is that the sky has levels 11-15 in it. But since ten is the brightest level that can be recorded, all those pixels were recorded at ten. Again, because there is no difference in the recorded brightness level, there is no detail in that part of the picture.

Ok, so we covered a lot today and actually started talking about histograms. But there is still a lot to cover including how to use this information when you are taking pictures and the importance of three-color-histograms. Like I said in the beginning, this is a complicated subject, but a very, very valuable tool. I hope you stick with it. Let me know if you have questions so far.

Read Part 4 next.